In Optical Transport Network (OTN), there is momentum regarding flex modulation and flex line rates. There are different ways to achieve a flex rate. For example, some vendors have created proprietary Optical channel Data Unit Group (ODUG) Super High Order (SHO) wrappers to handle cases of 50 G, 100 G, 200 G, etc. for modem technologies. In ITU-T, there have been discussions about defining OTUCn, a byte interleaved scheme for flexibility in increments of 100 G. OTUCn stands for Optical channel Transport Unit Cn where C means 100 and n is a multiplier of 100, e.g. OTUC2 is 200 GB/s, OTUC4 is 400 GB/s, etc. The current ITU-T OTUCn standards are planning on defining a modular (not necessarily flexible) architecture for 100 G slices and granularity. The problem is that this architecture does not give enough granularity on next-Gen devices for bandwidth versus performance/reach tradeoffs. It also does not cover some modulation rates (e.g., 8-Quadrature Amplitude Modulation (8QAM) at 150 G) that are not aligned to 100 G boundaries. Other initiatives have proposals to turn off single or groups of physical, virtual, or logical lanes in a Physical Medium Dependent (PMD) layer to achieve a desired rate. There are some significant implementation and logic complexities when designing a protocol to support multiple different rates at the physical layer.
This flexible line rate is becoming a hot topic in the industry and recent activities by end users include a desire for sub-100 G granularity (25 G or 50 G). Again, some conventional schemes address flexibility by turning off physical or virtual lanes, but complexity and logic cost is significant. Resizing using lanes scheme is also a challenge. The 25 G/50 G granularity does not line up well to existing 10 G traffic. Also, scaling conventional techniques for mux/mapping rates of 10 G up to 500 G requires large logic complexity.
As optical transmission systems start approaching the Shannon limit for non-linear noise and demand for increased data rates continues, Digital Signal Processing (DSP)/modem engines can get implemented in parallel devices or multiple engines get integrated to create super-channels with optical or electrical mixing. Low complexity and flexible schemes are needed for inverse multiplexing (“muxing”) and distributing signals across these different channels at the physical layer. To minimize line-side penalties, an equal and symmetrical bandwidth split is required across the multiple engines, and there is high complexity involved to support flexible rate bandwidth splitting. For example, to split 340 G across two devices would be 2×170 G channels, 930 G across three devices would be 3×310 G channels, etc.
There are different conventional techniques developed to handle the breakup and inverse muxing of signals across multiple channels. For example, IEEE has defined Link Aggregation Groups (LAG) and ITU has been using Virtual Concatenation (VCAT) type of schemes of standard defined containers Low Order (LO)/High Order (HO) Optical channel Transport Unit-k (OTUk). LAG is a higher layer protocol utilizing smaller-sized channels to carry a super-channel. The protocol is implemented at Layer 2 (Ethernet) and adds huge complexity and memory requirements. It is typically implemented using a Network Processing Unit (NPU) and other types of devices; LAG is not an appropriate approach to be integrated into optical DSP/modem devices. Standard ITU-defined VCAT schemes include grouping smaller sized standard containers, which would be Optical channel Data Unit-2 (ODU2) to get 10 G granularity on the line side. There is a large logic complexity to map a signal (i.e. 240 G) to nxODU2 (i.e. 24) and then switch and distribute these ODU2 signals across multiple optical DSP/modem devices. The extra mapping complexity can add to wander and decrease network performance.